Hi, We have to connect 32 dataloggers, separated 500m from each other (16km total) with an ethernet LAN. We're going to place 32 fiber optic switches close to each datalogger, and an utp link from switch to the datalogger. I think this network is excesively long due to ehernet inherent CSMA/CD timing considerations, and that we should split the network up using routers into smaller sub-networks. On the other hand, traffic is going to be quite low. Will it work like this or we should definitely add routers/bridges?
You are going to place one switch close to each datalogger, yes?
If it's full-duplex, you won't have collisions, and if the traffic is low enough to prevent overflowing buffers in your switches, you shouldn't lose any data.
If the long links are full duplex, then there is no CSMA/CD problem. Otherwise, for 10baseFL (10 megabit/s) you should be fine unless you daisy chain them.
since you are using switches, each fibre link is a separate CSMA/CD domain.
As long as the fibre links operate as full duplex you can support several Km per link.
Note that since you only want low capacity, you are better running at 100 Mbps rather than 1000. That way you can use multimode fibre 100 Base-FX interfaces.
If you use Gigabit then you either need 1000 Base-LX interfaces, or good quality fibre - multimode interfaces and fibre are limited to 260m or so
You will need to provide further details however:-
Assuming that the switches are:-
Full Duplex on the Fiber Optic ports at 10 or 100 or 1000 Mbps OR Are 10 Mbps
There there is no problem with CSMA/CD
The only way there could be an issue is with 100 HD where the max length is 200M (IIRC)
The standard (IEE 802.1d) does say however that the maximum network diameter is 8 bridges (bridge=switch pretty much). This though really applies to the Spanning Tree Protocol and I think that unless you need STP for redundant links then there will be no problem.
Even if you want to use STP I find it hard to imagine that it will not work since the standard was written years ago and took into account slower links and lower capacity switches.
I have never tried a network of > a few switches in diameter.
AH! Oh yes I have. Years ago my company used to test switches by configuring them with a lot of vlans with two ports in each. These VLANs were then connected by crossover cables making a very long network and a pair of PCs were used to ping through the lot. It worked. This was done with dozens of VLANs.
If you do want to stick to the standard you will need 4 routers to break the 32 nodes into 4 x 8 switch (approx) sections.
Ok, thanks. I assume the network consists on a virtual bus formed by full duplex, fiber optic switched links. So no colissions are possible. But if two loggers are trying to communicate at the same time in the same direction, _ There must be some sort of buffer in each switch in case both signals arrive at the same time, right? _ Even though the traffic won't be high at all, what would happen in case of buffer overrun? Would a colission signal find its way back to the sender datalogger? _ The datalogger must support full duplex or the switch could reproduce a colission signal if needed?
Yes, the switches do something along the lines of store and forward.
Again, if your traffic is low, you won't have buffer overrun. If packets are dropped for whatever reason, the sending datalogger won't receive it's TCP/IP ack, and will resend.
It's not a collision signal, it's a lack-of-ack that causes the retry.
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.